The Scintillating Grid Illusion

Transcription

1 Pergamon PH: S (96) Vision lk~,, Vol. 37, No. 8, pp , Elsevier Science Ltd. All rights reserved Printed in Great Britain /97 $ ()() The Scintillating Grid Illusion MICHAEL SCHRAUF,*~ BERND LINGELBACH,$ EUGENE R. WIST_f Received 30 November 1995; in revisedform ]9 Jllne IWj Disk-shaped luminance increments were added to the intersections of a Hermann grid consisting of medium grey bars on a black background. Illusory spots, darker than the background, were perceived as flashing within the white disks with each flick of the eye. This striking phenomenon may be referred to as the scintillating grid illusion. We determined the conditions necessary for canceling the Hermann grid illusion, as well as the luminance requirements and the size ratio between disks and bars that elicits the scintillation effect. The fact that scanning eye movements are necessary to produce the scintillation effect sets it apart from the Hermann grid illusion. ~ 1997 Elsevier Science Ltd. All rights reserved. Hermann grid illusion Scintillation effect Brightness perception Eye movcments INTRODUCTION The original Hermann grid illusion is characterised by light spots perceived at the intersections of a dark grid on a white background. These illusory spots were first reported by Brewster (1844) and Hermann (1870). Hering (1878) showed that dark illusory spots occur in a pattern of opposite contrast polarity (for a review see Spillmann, 1994). Baumgartner (1960, 1990) attributed this illusion to the differential stimulation of ON-or OFFcentre receptive fields, resulting in a net darkening or brightening, respectively. Troscianko (1982) measured the strength of the hollow Hermann grid illusion by locally increasing the luminance of the intersections until the grey spots could no longer be seen. The luminance required for cancellation provided a measure for the strength of the illusion. Bergen (1985) modified the standard Hermann grid by low-pass filtering. This operation resulted in a blurred grid whose intersections were more luminant than the bars. In such a grid, dark patches can be seen at the intersections during eye movement. This effect is the topic of this study. We first superimposed small uniform disks, increments or decrements, onto the intersections of a Hermann grid to cancel the illusory grey spots. As a result, we observed a striking phenomenon: scintillating dark spots within the white disks and scintillating light spots within the black disks (Fig. 1). These spots were perceived predominantly in peripheral vision, but can also be observed foveally with certain spatial conditions. As in the case of the Hermann grid illusion (Spillmann & Levine, 1971), the *To whom all correspondence should be addressed [Enrad schrauf@mail.rz.uni-duesseldorf.de]. ~Instituteof PhysiologicalPsychologyII, Heinrich-Heine-Universitat, Universitatsstr. 1, Dusseldorf, Germany. $Fachhochschulefiir Augenoptik,73430Aalen, Germany. illusory dark spots were stronger than the illusory light spots. In anticipation of more distinct effects we selected the dark version (Fig. 1, left) and then asked the following questions: first, what is the luminance of the disks required to cancel the illusory dark spots in the Hermann grid for various combinations of bar luminance and background luminance? Second, what is the relationship between the rated strength of the dark spots in the scintillating grid illusion on the one hand and the luminance of the disks, the bars, and the background, on the other, when two parameters are kept constant? I.,ast, how does rated strength depend on disk size and bar width? METHODS Fifteen students, nine females and six males, who were naive as to the purpose of the experiment served as subjects. They had normal or corrected-to-normal visual acuity and normal contrast sensitivity. Grid patterns representing a matrix of 8 x 6 intersections were generated by an IBM 80486/50 computer (frame rate 100 Hz, 256 grey levels) and displayed on the screen of an EIZO 17 RGB monitor (Model T560i-T, Sony Trinitron tube). Luminance were measured with a Minolta Luminance Meter (Model LS 100). The subject s head was supported by a chin rest located 70 cm away from the monitor. Unless specified otherwise, the angular dimensions of the stimulus were as follows: bar width 0.31 deg, background square width 1.54 deg, and disk diameter 0.43 deg. The ratio between the diameter of the disks and the width of the bars was chosen on the basis of a pilot study and was 1.4:1. Experiments were performed with free viewing of the stimulus patterns and there was no time limit for the subjects to respond. 1033

2 1034 M. SCHRAUFet al. FIGURE1.Scintillationeffect. Dark illusoryspotsare perceivedwithin the white disks (left) and light illusoryspotswithin the black disks(right). These spotsare seenbest in the peripherywherethey blink,or scintillate,with each eye movement.note that the spots can also be seen foveally when the observationdistance is increased. The scintillation effect disappears with steady fixation. These patterns are shown here for demonstrationonly and are not identical with the stimulus patterns used in the experiments. Experiment 1: cancellationof the Her-manngrid illusion Using a matching task, the subjects were asked to adjust the luminance of the disks until the illusion was cancelled. There were seven Iuminances for the bars, rangingfrom 0.42 to 42.5 cd/m2and three luminance for the background, ranging from 0.03 to 6.77 cd/m2. This resulted in Michelson contrasts varying from 0.25 to Experimentalconditionswere randomizedfor each subject acrossbar luminanceand backgroundluminance. Four conditions where disk luminance was below bar luminance were not tested. The resulting 17 conditions were repeated three times. Figure 2 showscancellationluminanceof the disksas a function of bar luminance for three background lumi- Bar Luminance[cd/mq FIGURE 2. The disk luminance required for canceling the illusory spots in the Hermann grid at different background luminance is plottedas a functionofbar luminance.in this andthe followingfigures, data points are averages of 45 individualratings of 15subjects. The vertical bars indicate + 1 standard error. nances. Data points in this and subsequent figures are averages of 45 individual settings. Values of mean canceling disk luminance specify combinations of bar and background luminance at which the Hermann grid illusion can just be cancelled by appropriate disk luminance. At higher bar and lower background luminance the canceling disk luminance is proportionately higher than at lower bar luminance; that is, the contrast between bar and background luminance must be proportionatelygreater. This implies that the strength of the Hermann grid is proportionatelygreater at higher bar luminance as well as greater bar/backgroundluminance difference. The scintillation effect occurred only when the luminance adjustments for the disks were above the luminance required for nulling. In the following experiments we studied the requirementsfor this effect. Experiment2: the scintillationeffect as ajimction of disk luminance For a given luminanceof the bars (11.5 cd/m2)and the background(1.27 cd/m2),we presented disks of different luminance ranging from 0.42 to 142 cd/m2. Spatial parameters were held constant (bars 0.31 deg, background 1.54 deg, disks 0.43 deg). Owing to the fact that the Hermann grid illusion is easily eliminated by small increases of disk luminance above the bar luminance (Troscianko, 1982), we used 11 fine shades of disk luminance for the Hermann grid illusionand eight more coarse shades for the scintillation effect. The subjects were askedto rate the strength of the illusorydark spots for each grid. Specifically,they were to use a rating scale on which a value of l would indicate no illusion, ratings of 2 to 4 would indicate an illusion that was

3 THE SCINTILLATINGGRID ILLUSION 1035 I I Hermann Illusion Disk Luminance [cdlm~ FIGURE 3. Mean rated strength of the Hermann grid illusion (descending branch on left) and the scintillation effect (ascending branch to the right) is plotted as a function of disk luminance. The maximumhermanngrid illusion (rating of 3) occurs when no disk is superimposed onto the intersections. Bar luminance, 11.5cd/m2; backgroundluminance, 1.27cd/m2.The maximumscintillation effect (rating above4) occurswhenthe luminanceof the diskis a factor of 10 above that of the bars. The vertical bars indicate + 1 standard error. stronger (darker, or more numerous remained open, we did not attempt to avoid ratingsbased upon the perceived number of illusory spots rather than the perceived individual strength), and 5 would be allowed if the illusion they expected from a particular imagined grid was found to be optimal in a preliminary study for some subjects, including the authors. Following these general points the subjects were asked, according to the rating scale, to assign a number to the perceived illusory strength. Disk luminance were varied randomly but separately for Hermann grid illusionand the scintillationeffect. For each disk luminance three measurements were made consecutively.this procedure was adopted in all further experiments. Figure 3 shows the mean rated strength of the grey spots at the intersections.in effect, Fig. 3 is actually two figures: one shows the reduction in strength of the Hermann grid illusion as a function of increased disk luminance (left hand side, marked as Hermann Illusion ),while the other depictsthe increasein the strength of the scintillation effect (right-hand side, marked as Scintillation Effect ) with further increasing disk luminance. Experiment 3: variation of bar luminance To determine the effect of bar luminance, test stimuli were selected for a disk luminance that produced the maximal scintillationeffect (Fig. 3) and a bar luminance that permitted a range of responseswithoutencountering ceiling or flooreffects.thus, for a given luminanceof the disks (142 cd/m2) and the background (1.27 cd/m2),we randomly presented bars with 12 different luminance ranging from 1.5 to 30 cd/m2. The subjects and rating procedure were the same as in Experiment 2. E,,,, Bar Luminanca [cd/m*] FIGURE4. Meanrated strengthof the scintillationeffect is plottedas a function of bar luminance. Disk luminance, 142cd/m2;background luminance, 1.27cd/m2.The vertical bars indicate L 1 standard error. Figure 4 shows the results.the mean rated strengthof the scintillating spots first increases rapidly with increasing bar luminance up to a maximum rating of about4, at which pointbar luminance are approximately seventimes greaterthan the backgroundluminance.with a fuither increaseof bar luminancethe scintillationeffect decreases. Experiment 4: variationof backgroundluminance Conditions for a maximum scintillation effect determined in Experiments2 and 3 (Figs3 and 4) were used to study the effect of variation in background luminance. For a given luminance of the disks (142 cd/m2)and the bars,(11.5 cd/m2), we randomly presented backgrounds with 11 different luminance ranging from 0.03 to 11.5 cd/m2. Res@s Figure5 shows the mean rated strength plotted as a function of background luminance. The curve decreases I 1 I o Background Luminance [cd/mz] FIGIJRE5. Meanrated strengthofthe scintillationeffect is plottedas a function of background luminance. Disk luminance, 142cd/m2; bar luminance, 11.5cd/m2.The vertical bars indicate f 1 standard error.

4 1036 M. SCHRAUFet al. 5 Mean Rated Strength FIGURE6. Meanrated strengthof the scintillationeffect is plottedas a function of disk size with bar width as parameter. Background luminance was 0.03 cd/m2. monotonically and reaches a magnitude of 1 (no effect) when the luminance of the backgroundand bar are equal (i.e., no grid present).a pronouncedscintillationeffect is observed only with the lowest backgroundluminance. Experiment 5: variation of disk size and bar width In this experiment the strength of the scintillation effect was measured as a function of the spatial parameters of the stimulus pattern. There were ten disk sizes (ranging from 0.06 to 0.6 deg) and ten bar widths (ranging from 0.06 to 0.6 deg), while background size was kept constant (1.54 deg). Disk luminance was 142 cd/m2, bar luminance was 11.5 cd/m2, and background luminance was 0.03 cd/m2. Figure 6 shows that for a given bar width, the mean rated strength of the scintillation effect first increases with increasing disk size, reaches a peak and thereafter decreases. Rated strength is maximum when the ratio between disk size and bar width is approximately 1.4:1 (range 1.2:1 to 2:1). With increasing bar width, curves start out shallow indicatingthat the scintillationeffect is virtuallyabsentfor disk sizes smallerthan the bar. Within the range of bar widths used, a peak was reached only for the smaller widths. SUMMARYOF RESULTS The resultsobtainedin this studyshowthatthe strength of the Hermann grid illusion varies with bar and backgroundluminance (Fig. 2). For a given combination of bar and background luminance the Hermann grid illusion shifts over to the scintillation effect (Fig. 3). When the disk luminance was 12 times the background luminance, the intensity of the scintillation effect exceeded greatly that of the Hermann grid illusion. Furthermore, for combinations of disk and background luminance or disk and bar luminance, the scintillation effect could only be seen within a rather narrow range of stimulus parameters. The effect was strongestwhen the bar luminance was about seven times greater than the backgroundluminance(fig. 4) and when the background luminance was low (Fig. 5). At the same time, the ratio between disk size and bar width had to be about 1.4:1 (Fig. 6). DISCUSSION Our results indicate at least three prerequisitesfor the scintillationeffect: 1. A grid capable of eliciting the perception of the classicalhermann grid illusionmustbe present. 2. Luminanceincrementsor decrementswhich exceed those required to cancel the Hermann grid illusion must be superimposedon the intersections. 3. Stimulationmust be brief. With steady fixation the scintillationeffect subsidesand quickly disappears. Voluntary scanning eye movements were used to produce brief stimulation.the question of whether such scanningeye movementsare necessaryor only sufficientremains open. Concerningthe firstprerequisite,it might be suggested that the scintillation effect could be accounted for in terms of a Mexican hat model with single small concentric receptive fieldsproducing more lateral inhibition at the crossingsthan anywhere else (Baumgartner, 1960, 1990). As stated above, a grid capable of producing the illusoryspotsof the Hermann grid is a prerequisitefor the effect; hence lateral inhibitionmechanisms are involved and obviously necessary but not sufficient. One could assume after examining Fig. 1 that an even larger concentric receptive field with a centre correspondingin size to that of a white disk and centred on one of them would be sufficient to account for the effect. This receptive field would have to possess a large surround encompassingneighboring disks. On the basis of such a model, one would expect the scintillation effect to be present in Figs 7(a, b). This is not the case, as can be seen by comparing Figs 1 and 7. Both fulfil the requirements of such a receptive field, but the scintillation effect is observedonly in Fig. 1,where the disksare superimposed on the intersections. Preliminary observations suggest that a minimumof 3 x 3 evenly spaced intersectionswith superimposed spots is required to produce the effect. Thus, a model which fails to take into account the distributionof the disks in relation to the intersectionsis not sufficient.a more complexreceptivefieldresponsive to this distribution as well as to eye movements and sensitive to orientationwould be required. Furthermore, since a minimum of elements is required the illusion does not occurwith an isolatedintersection,as in the case of the Hermann grid (Wolfe, 1984)-the effect does not fit in well with a model involving purely local inhibitory and excitatory interactions. The requirement of a minimum number of orderly arranged elements suggeststhe participation of global, in addition to local, processes. Von der Heydtet al. (1991)found cells in areas V1 and V2 in the monkey which respondedto gratingsand other periodicpatterns but not to single bars and edges. They

5 a) b) THE SCINTILLATINGGRID ILLUSION 1037 FIGURE 7. The scintillation effect becomes weak or even absent in some spatial variations. regarded their observations as being incompatible with linear filtermechanisms.these propertiesare compatible with our observationthat a periodicpattern is requiredfor the scintillation effect: multiple, evenly spaced disks located at corresponding intersections. Furthermore this observation may indicate an involvement of global cortical mechanismsof the kind proposed for the linking and groupingof features acrossdistance(eckhorn,1991). Eckhorn et al. (1992) regard the linking field as enabling the receptive field properties of neurones in different parts of the visual field to be linked into perceptual wholes. An interesting property of such linking fields is that they are transiently constructed by the neurones involved in the co-operative process. This property fits in nicely with the scintillatingcharacter of the illusion. The considerations of Spillmann and Ehrenstein (1996) of global processes in relation to the neuronalbasis of Gestaltphenomenaalso agree well with this line of reasoning. We have not as yet systematically investigated the influence of temporal factors on the illusion. Brief exposures of the Hermann grid in the dark adapted eye eliminate the dark spots (Wist, 1976).If the temporal, as well as the spatial conditionsfor producingthe Hermann grid illusion are necessary requirementsfor the scintillation effect, then dark adaptation should eliminate it as well. Furthermore, it must be determined whether the effect of scanning eye movements reduces the necessity of stimulating receptive fields briefly. preliminary experiments in which the effect of slow pursuit eye movementswere investigatedindicate a weaker scintillation effect as compared to saccadic eye movements. In this context, saccadic omissionor suppression(campbell & Wurtz, 1978; Corfield et al., 1978) may also play a role. Furthermore, neurophysiologicalexperiments extending the work of Schepelmann et al. (1967) on the Hermann grid to the scintillation effect could provide more concrete information concerning the nature of the underlyingreceptivefieldorganization.using a Hermann grid, they recordedfrom singlevisual cortical cells in the cat, whose firing rate was reduced when stimulated simultaneously with horizontal and vertical bars, as opposed to stimulation with either bar alone. They interpreted this decrease in firing rate as the neural correlate of the darkening at the intersection. If the receptive field characteristics required for the scintillation effect are similar to those of the Hermann grid illusion,then the positioningof a disk at the intersections whose luminanceis at least twice that of the bars should result in an even greater reduction of firing rate. The predictionbased on the resultsof the present experiments would be that this would not occur, since only local processingwould be involved. CONCLUSION The fact that stimulus conditions resulting in the Hermann grid illusionare necessarybut not sufficientfor producing the scintillation effect implies that a neurophysiologicalaccountof it must go beyond one based on lateral inhibitionwithin single receptive fields. Both the conditionsnecessaryfor producingthe scintillationeffect beyond those required for the Hermann grid and its uniqueperceptualqualityjustify regardingit as a separate perceptual phenomenon. REFERENCES Baumgartner,G. (1960). Indirekte Gro13enbestimmungder rezeptiven Felder der Retina beim Menschen mittels der Hermannschen Gittertauschung (abstract). P@gersArchivfir die gesamte Physiologic, 272, Baumgartner,G. (1990}Where dovisual signalsbecome a perception?

6 1038 M. SCHRAUFet al. In Eccles J. & Creutzfeldt O. (Eds), Theprinciples of design and operationof the brain(pp ).Berlin, Heidelberg,New York: Springer. Bergen, J. R. (1985). Hermam s grid: new and improved (abstract). Investigative Ophthalmologyand VisualScience, Supplement26, 280. Brewster, D. (1844) A notice explaining the cause of an optical phenomenonobservedby the Rev. W. Selwyn.In Murray,J. (Ed.), Report of the FourteenthMeeting of thebritishassociationfor the Advancementof Science (p. 8). London. Campbell, F. W. & Wurtz, R. H. (1978). Saccadic omission:whywe do not see a gray-out during a saccadic eye movement. Vision Research 18, Corfield, R., Frosdick, J. P. & Campbell, F. W. (1978). Gray-out elimination: the roles of spatiaf waveform, frequency and phase. ViswnResearch, 18, Eckhom R. (1991) Stimulus-evoked synchronization in the visual cortex: linking of Iocaf featores into globaffigures? In Kriiger, J. (Ed), Neural cooperativity (pp ). Berlin, Heidelberg, New York: Springer. Eckhom, R., Schanze, T., Brosch, M., Salem, W. & Bauer, R. (1992) Stimulus-specific synchronization in cat visual cortex: multiple microelectrodeand correlationstudiesfrom several cortical areas. In Basar, E. & Bullock,T. (Eds),Inducedrhythmsinthebrain(pp.48 80). Boston, Basel, Berlin: Birkhauser. Hermann, L. (1870). Eine Erscheinung simultanen Kontrastes. PjJiigersArchivfir die gesamtephysiologic,3, Hering E. (1878)ZurLehre vomlichtsinne,gerold, Wien. Schepelmann, F., Aschayeri, H. & Baumgartner, G. (1967). Die Reaktionender simple field -Neurone in Area 17der Katze beim Herrnann-Gitter-Kontrast(abstract).PjhlgersArchivfiir die gesamte Physiologic,294, R57. Spillmann,L. (1994). The Hermann grid illusion: a tool for studying humanperceptive field organization.perceptio~ 23, Spillmann,L. & Ehrenstein, W. H. (1996) From neuron to Gestalt: mechanisms of visuaf perception. In Greger, R. & Windhorst, U. (Eds), Comprehensivehumanphysiology, Vol. 1 (pp ). Berlin, Heidelberg: Springer. Spillmann, L. & Levine, J. (1971). Contrast enhancement in a Hermanngrid withvariable figure-groundratio.experimentalbrain Research 13, Troscianko,T. (1982).A givenvisual fieldlocationhas a wide range of perceptive field sizes. VisionResearc~ 22, von der Heydt, R., Peterhana, E. & Diirsteler, M. R. (1991) Grating cells in monkey visual cortex: coding texture? In Blum, B. (Ed), Channelsin the visual nervous system: neurophysiology,psychophysics andmodels (pp ). Landon, Tel Aviv: Freund, Wist, E. R. (1976). Dark adaptation in the Hermann grid illusion. PerceptionandPsychophysics,20, Wolfe, J. M. (1984). Global factors in the Herrnann grid illusion. Perception,13, Acknowledgements Thispaper represents a part of the Habilitation thesis of the first author.the authorsare indebtedto Lothar Spillmann for his invaluablecritical readingof, andcommentson,the manuscript. Thanksare also due to Walter EhrensteinandJohn S. Werner for their valuable comments. We thank the reviewers for constructive commentswhich contributedsubstantially toward clarifying the text.